|Anonymous | Login | Signup for a new account||2020-02-27 13:15 GMT|
|Main | My View | View Issues | Change Log | Roadmap|
|View Issue Details|
|ID||Project||Category||View Status||Date Submitted||Last Update|
|0001385||SiT!||incidents||public||2010-08-21 17:50||2012-01-24 13:21|
|Target Version||Fixed in Version|
|Summary||0001385: Auto Assign lottery should have a 'fair' mode|
|Description||Its been requested we should have a fair distribution algorithm e.g.|
If you have 3 users with no incidents and 9 new incidents are added they should all get 3 each, at the moment its not 'fair'
(Requested by legoman)
|Tags||No tags attached.|
I'd like to see an explanation of why the current lottery isn't fair. There is nothing that makes the lottery have any kind of bias, it's random but weighted according to a number of fairly sensible factors (see http://sitracker.org/wiki/Auto_assign [^])
With 9 incidents and three engineers with the same skills, same status and no incident already in their queues more often than not the distribution will be three incidents each. There is some randomness involved though so sometimes one engineer may get 4 incidents while the others get 3 and 2 respectively but I don't think that makes this an unfair way to decide who gets incidents. Each engineer had equal chance of getting 4 or 2.
I suppose whats being proposed here is to remove the randomness... a reasonable request on first glance, but in order to be realistic we need a proposal of how to make it fair without being random.
This discussion has gone on a while here in our support team, as some guys thought it was "not fair enough", and even after proving to them how the lottery system worked, the final request from about half of the team was to be able "switch off" the lottery system, and in this case to assign the cases as one would deal cards .. one for you one for me type of thing ..
Thus in your example of 3 engineers (that are of the same group of course), the first incident goes to engineer one, the second to engineer 2, then engineer 3, and then we restart at engineer 1. If engineer 2 is on holiday, we skip him; If we last assigned to engineer 2, then next case that arrives tomorrow morning goes to engineer 3.
Personally i like the lottery system, because it is not so easy to "cheat" with it :) and in the end after some time it seems to me the results are quite equal.
Ivan, this was a request from legoman, given his implementation they need to given an even number of 'incidents' to each engineer. Whilst overtime I'm sure it will balance out though in the short term some people get more incidents.
I think an option to turn the randomness off would be useful - or perhaps a more plugable/extensible lottery algorithm system
|Confirming as an idea for a pluggable lottery algorithm. Not sure how we can improve the "fairness" of the current algorithm without concrete suggestions of a new algorithm.|
|All these issues are discussed in 0001212|
|2010-08-21 17:50||paulh||New Issue|
|2010-08-22 18:57||ivan||Note Added: 0003369|
|2010-08-23 08:42||nicdev||Note Added: 0003370|
|2010-09-21 21:05||paulh||Note Added: 0003415|
|2012-01-24 13:09||ivan||Note Added: 0004422|
|2012-01-24 13:09||ivan||Status||new => confirmed|
|2012-01-24 13:19||ivan||Relationship added||related to 0001212|
|2012-01-24 13:21||ivan||Relationship replaced||duplicate of 0001212|
|2012-01-24 13:21||ivan||Note Added: 0004427|
|2012-01-24 13:21||ivan||Status||confirmed => closed|
|2012-01-24 13:21||ivan||Resolution||open => duplicate|
|Copyright © 2000 - 2020 MantisBT Team|